-
Notifications
You must be signed in to change notification settings - Fork 79
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
is this better than diff svc and so vit svc? #17
Comments
whisper is from openai . |
mm.. does that mean it's better? |
from now on, whisper is the best audio encoder for svc.这个项目的目的就是为了去证明这一点。 |
您有对比几版相同模型不同content提取方式的结果吗?比如hubert,whisper,content vec,或者类似so vit svc的第九层的content vec就用您现在的方案。 |
@hongwen-sun 你我的目标不同,您期望的是一个音质效果好的SVC;我是在做技术方案的研究,比如whisper用于歌声转换的有效性,还有说话人自适应;我需要一个高效的方案来验证我的一些想法。 |
此外,我个人比较崇尚简洁。您如何看待 ‘奥卡姆剃刀原理’ 呢? |
我感觉简洁也可以达到比较好的效果,并不冲突,效果的差距通过block结构的简单修改和参数量的增加完全可以弥补。 |
说白了,我本来是想白嫖您的结论,看看如果您做了对比,哪个好一点 |
whisper是目前为止能获取到的效果最好的多语言ASR模型,它的好坏由他的识别率就可以直接体现,它通过海量的多语言数据训练得到,这是其他开源模型无法媲美的。如果您要做语音转换可能hubert等其他自监督模型比较合适,它们包含了其他非语言的信息、如语气&情感等,也许还有性别信息、以及可能的泄漏的音色; |
好的,非常感谢。 |
@MaxMax2016 你好,我做了个对比的实验,使用不同ppg训练声学模型的情况。训练集的loss如下: whisper:medium |
@hongwen-sun 赞,有疑惑然后去验证,是个搞科研的好手! |
@MaxMax2016 您好,可以加一下微信吗,可以做一些交流? |
Btw, there is argumentation behind this that whisper was actually trained on songs. It is very good at speech recognition of songs compared to many other speech models. |
I would like to know if this is better than diff svc and so vit svc
The text was updated successfully, but these errors were encountered: